Flip Display Clock Project

[Mysion]

I wanted a fun project to try and complete before school starts back up in the fall. Thankfully I have a few old flip displays on the "from ebay" pile of electronics. And what a pile it is.

If I can I want to build a slick wood case for it. I do have limited wood working skills and a CNC to call on. I thought I'd document it all here and get feed back on some parts. I've mostly finished the electornics design portion of the clock.

Obviously I need a driver for the flip display. I'll just use a few discrete transistors for that. If I can I want to have a sun rise alarm feature. This means lighting up the room to simulate sun rise. I figure a few LED's will work.

Enough rambling  here's the star of the build a 7 segment flip display.

9$ a pop on ebay. I got five of them. My major regret is that I did not get more.


Here's the current schematic. There's an RTC for obvious reasons(2% ppm) and a I2C GPIO exapnder since I ran out of pins. I'm using an ATMega 48A with every single pin in use (except two analog pins)



This is the real star of the show, the driving circuits for the display. The section on the bottom supply's 12v and selects the segment. The upper portion allows you to select the direction and the segment to flip. This was more complex than I first expected. It might be difficult to fit this on the board... Oh well shouldn't be too bad. That many discrete transistors on will make it look interesting.


Next I need to start working on the PCB design and thinking about the physical construction. I'd rather not have a hot glue based clock so I need to give that some real thought.

So what do you guys think so far?

[tooki]

Looks cool! Comment to subscribe, since I'm working on a clock project myself.

Is that the DS3231 RTC?

[Mysion]

Yep DS3231 is the chip! I love that RTC. Expensive as all heck but no more than a minute deviation a year. No fussing around with IOT and time servers needed. 

What clock project are you working on?

[Gyro]

You Might want to check out Marco Reps's electromagnetic clock project.



Check out the minimum segment switching demo at about 1:30 

[Benta]

Way cool!
I remember having one of those displays in my hand many years ago, and wondering at how light the individual segments are and how smooth the action is.

The only step up would be using those "rolodex-like" digits they still have for the big departure displays at Frankfurt/Main airport (Germany):


A bit worn, and something in the synchronization is failing, though.

[128er]

Followed!

And here is my flip clock. Driver PCB is still prototype. The biggest challenge in this project are my horrific coding skills ^^

[NivagSwerdna]

OOOOO those displays are lovely... if you see a similar set on ebay please post the link.

[tooki]

Yep DS3231 is the chip! I love that RTC. Expensive as all heck but no more than a minute deviation a year. No fussing around with IOT and time servers needed. 

What clock project are you working on?

Like $2? I don't consider that too bad. Certainly worth the added cost over the crappy 1302/1307 RTCs!

The project is fundamentally a giant-digit alarm clock, something I can see without putting on my eyeglasses. With my eyesight, that means digits at least 3"/10cm tall for a clock at arm's length.

Of course, large digit clocks exist for sale, but other than big digits, they tend to be shit (bad user interfaces, etc). So I wanna make one that's beautifully designed in every other way, too! Of course, the building is the main aspect, and it's a good way to put my spotty knowledge to the test -- it's been a great learning experience. (Some of the things I've needed to learn are: driving displays of various types, I2C and SPI, RTCs, buttons and matrices, and various aspects of coding*.)

I'm still not sure what kind of display I'm gonna use. I've been tinkering with a 64x32 RGB panel, but an Arduino struggles to drive it (it is 6,144 LEDs, after all!), and can't do high bit depths, the MCU just isn't fast enough. Another option is gigantic 7-segment LED displays (I finally found a source for white ones, all the way to 5"/12cm tall!). Making 7-segment displays out of WS8212B LED strips is another, which would allow full RGB without too much CPU load. I've even considered making 7-segments from LED filaments, though this is far harder to drive, since they're designed for 60+ volts. (The only firm requirement, aside from size, is that it not be a red LED display. I have an irrational hate of red LED displays.)

Obviously, a display that big, which is bright enough to be visible by day, would be insanely bright at night, so I'm gonna have to use ambient light sensing to adjust it for room light.

Recreating vintage display technology is of course also interesting. Aside from making mock-Panaplex displays with the LED filaments, two other fun approaches would be side-lit laser-etched plastic sheets with the segments, or projection displays.

Other pending items include: generating a pleasant alarm sound, maybe using a GPS or DCF receiver module to receive accurate time signals, perhaps a proximity sensor to wake up the display from dimmed by waving a hand nearby, and perhaps in the future interfacing with lighting projects.

* I'm a terrible programmer, and the coding has been the biggest challenge -- still pending are elegant button handling, and date operations. Real OSes have calendar libraries (with math functions, like adding/subtracting months/days/hours/mins) that handle things like leap years automatically, but I haven't found anything like that for Arduino, and I know that date handling is one of those "Here Be Dragons" minefields full of edge cases...

[NivagSwerdna]

Try RTCLib

[tooki]

Try RTCLib

I've seen it before. It's helpful, but still far too rudimentary (and poorly documented!).

For example, take the task of setting a calendar by incrementing/decrementing a month. It's easy if you're mid-month:

2017-08-04 + 1 month = 2017-09-04 - OK
2017-08-04 - 1 month = 2017-07-04 - OK

But what if you're at the end of the month? What do you do with a remainder?
2017-06-30 + 1 month = 2017-07-30 - OK
2017-06-30 - 1 month = 2017-05-30 - OK
2017-07-31 - 1 month = 2017-06-30 - not simple (naive calculation would produce July 1)
2017-05-31 + 1 month = 2017-06-30 - not simple (naive calculation could produce July 1)

Additionally, from a UI perspective, even if you've handled the cases above, you want there to be "memory" as you increment/decrement. For example, suppose you're decrementing by month from July 31. You want it to remember day "31", even as you click through months with fewer days:
July 31 -> June 30 -> May 31 -> April 30 -> Mar 31 -> Feb 28 -> Jan 31 -> Dec 31 (y-1) -- good
July 31 -> June 30 -> May 30 -> April 30 -> Mar 30 -> Feb 28 -> Jan 28 -> Dec 28 (y-1) -- bad

(This is analogous to how the cursor on a computer remembers where it was on a line as you use arrow keys to move up and down, even passing through lines that are much shorter, but returning to the original horizontal position when possible, until reset by acting on the horizontal position by editing text or using the left/right arrow keys.)

Months are really the problem child of date calculations, since they're of variable length, so you can't simply add/subtract a fixed number of seconds from UNIX time.

[Mysion]

Gyro

I will have to take a look at that. Looks nicely documented. Very well done video.
I like the look he has with the flipping order. I concern I have with that is I have observed that some segments like to flip to a certain position if shook slightly.
Maybe it'll be different when I drive is magnetically.


Benta

I love those displays too! Maybe the next version. 
I've been to a train station in New York with those and the sound is very pleasing if loud.


128er

Looking good! Do yours have a center tap? I've seen the same ones you've got on ebay and no center tap would make driving much more difficult. You've got 4 clocks worth ya lucky bastard! 


NivagSwerdna

Sure here's a link to a similar set  http://www.ebay.com/itm/7-segment-1-Ferranti-Packard-electromechanical-display/160945344354?ssPageName=STRK%3AMEBIDX%3AIT&_trksid=p2057872.m2749.l2649


Tooki

I think we are talking about different variants that fall under the same name. I'm talking about this chip https://www.digikey.com/product-detail/en/maxim-integrated/DS3231S-T-R/DS3231S-T-RDKR-ND/4895475
It's only off by a minute a year at worst. There might be similar chips out there for less but it may be too late to change at this point.... 

I know what you mean by crappy clocks! Good ones cost too much and I'm a cheap bastard!  Large digits are a challenge. I did see some large flip ones on ebay but at 80$ a pop... At sizes that big you could make your own segments out of a diffuser if you wanted. Large IN-18 nixes might work too.

As far as accurate time goes I'm using a chip that's off by a minute a year at WORST. 8$ a chip but simpler than GPS. The last thing I want is another clock that has to be adjusted multiple times a year. I have a enough of those. It has one job and it has to do it well!

As far as calender's go an RTC chip handles it all automatically. Atmel has a app note on implementing a calendar. I can find it if you want. Good luck! Maybe I'll give you some good ideas.

[tooki]

I think we are talking about different variants that fall under the same name. I'm talking about this chip https://www.digikey.com/product-detail/en/maxim-integrated/DS3231S-T-R/DS3231S-T-RDKR-ND/4895475
It's only off by a minute a year at worst. There might be similar chips out there for less but it may be too late to change at this point.... 

I think you misunderstand; I was saying that I don't think $2 for a DS3231 is unreasonable. I was saying that it's worth paying extra for a DS3231, because the cheaper DS1307 and DS1302 RTCs can be very inaccurate. I'm well familiar with the DS3231, since it's what I am using in my project, too.

I know what you mean by crappy clocks! Good ones cost too much and I'm a cheap bastard!  Large digits are a challenge. I did see some large flip ones on ebay but at 80$ a pop... At sizes that big you could make your own segments out of a diffuser if you wanted. Large IN-18 nixes might work too.

Though they're sexy as hell, the digits on the IN-18 are only 40mm (1.8") tall, nowhere near big enough for my purposes.

As far as accurate time goes I'm using a chip that's off by a minute a year at WORST. 8$ a chip but simpler than GPS. The last thing I want is another clock that has to be adjusted multiple times a year. I have a enough of those. It has one job and it has to do it well!

Yep, the DS3231 is very accurate. $8?!? They're a dollar on eBay, premounted on a little breakout board.

As far as calender's go an RTC chip handles it all automatically. Atmel has a app note on implementing a calendar. I can find it if you want. Good luck! Maybe I'll give you some good ideas.

RTCs handle timekeeping (including dates) automatically. But they don't handle anything else. Since they don't have commands to increment/decrement (they only let you write the time/date to the registers, not perform operations on them), then you need to do those yourself, as I illustrated above. And that means implementing all the logic to produce valid dates to push into the RTC.

[128er]

128er
Looking good! Do yours have a center tap? I've seen the same ones you've got on ebay and no center tap would make driving much more difficult. You've got 4 clocks worth ya lucky bastard! 

No, no center tap on the coils. Each segment coil is individual. If you mean that. <- Edited my brain fart

I have enough 7 segment displays to build ten clocks. I bought three complete displays from gas pumps and desoldered all the individual 7 segment displays. I drive them in a matrix with NXP MC33880 driver.

[Mysion]

Tooki

I have a few of the ebay versions but I'm skeptical if they are real or not.  Mouser has a similar price to digikey. The chip is very slick in the accuracy department but you would be hard pressed to determine if one is a counterfeit.  I'm only making one clock so I really do no want to risk a fake chip that's just a normal RTC at heart. Heck whats 7 more $ after spending 80$ for the PCB and buying good wood. I'll do it for the piece of mind.

RTC handling on the micro's side isn't too bad. 99% of the time you won't hit the edge case of changing the day right before leap year. Depends on if your building for your self or production. It can be weird though.

Take a look at page 7 of this app note. http://www.atmel.com/images/atmel-1259-real-time-clock-rtc-using-the-asynchronous-timer_ap-note_avr134.pdf


There's a really nice flow chart there that shows all the things to check for.


128er

Nice number of displays! If you play your cards right you'll have wedding gifts for decades!

[Mysion]

The PCB was fun to work on. The limiting factor was the size of all the switches and buttons on the back. I only have 6 inches to work with and it came out to about 5.4 inches. They also take up most of the space on the PCB. Fitting the driver circuit in was tight. Not as tight as my BCD decoder but it was "interesting" 



On the lower left you have the 12v module to generate the power to drive the displays. On the lower right you have the coin cell and the RTC. The RTC is under the coin cell. With such a large ground plane it's the perfect location. In the lower middle you have the driver circuitry happily taking up both sides of the board.

The buttons take up the top portion. I could have used SMD buttons but I trust though hole more for that. The end total is 8.6 square inches and 321 connections.

Here's the board in glorious 3D



I think I'll use OSH parks fast service. It's pricey but i REALLY want this done before school starts. I think this is one of my largest PCB's to date. Let's hope I don't need a version 1.1

[tooki]

Tooki

I have a few of the ebay versions but I'm skeptical if they are real or not.  Mouser has a similar price to digikey. The chip is very slick in the accuracy department but you would be hard pressed to determine if one is a counterfeit.  I'm only making one clock so I really do no want to risk a fake chip that's just a normal RTC at heart. Heck whats 7 more $ after spending 80$ for the PCB and buying good wood. I'll do it for the piece of mind.

RTC handling on the micro's side isn't too bad. 99% of the time you won't hit the edge case of changing the day right before leap year. Depends on if your building for your self or production. It can be weird though.

Take a look at page 7 of this app note. http://www.atmel.com/images/atmel-1259-real-time-clock-rtc-using-the-asynchronous-timer_ap-note_avr134.pdf


There's a really nice flow chart there that shows all the things to check for.

Well, you coudn't really pass off a lesser RTC as the DS3231, since they use external crystals, while the DS3231 uses an internal temperature-controlled oscillator. The DS1307, for example, is neither code- nor pin-compatible, being just an 8-pin device that does less.

That flowchart is useful indeed, thank you for that!

As for doing it right: one of the things that annoys me most about most clocks for sale is how bad the user interfaces are. So that's definitely not somewhere I'm gonna skimp just because it's a project for myself. On the contrary, it's precisely because I value good user interfaces that I want it to work well, not just OK, never mind something silly like needing to be connected to a computer to set the time (like many DIY clocks I've seen do).

Regarding your PCB: where are the displays mounted?

[Mysion]

I'm still skeptical  after the whole fake FTDI chip thing. They could easily stick a simple crystal in or run it with a micro like the ftdi chips. You would never know it's not a TCXO until it's too late. Maybe I'm being slightly paranoid. 


I think you'll be slightly disaponted with my user interface then lol. I'm just sticking to 4 buttons to set the alarm time and clock time. Then a switch to turn the alarm on and a switch for the sunrise alarm feature.

You could make a really good one with a small LCD and a miniature joystick. My current monitor uses a small joy stick and it makes a world of difference.

My displays will be mounted on a separate frame with 19 wires plugging into a header on the PCB. I'll be quite a job connecting all of them but better that then pay for such a huge PCB!

[Mysion]

This is the 3D model for the subframe or frame that will hold the displays. I went though a few iterations of it making minor adjustments. There will be a few more changes done to it yet. The far right 5th display will be used for AM/PM and alarm indicators. You can see the gaps for wires at the bottom The bottom gap will also be used to make space for the PCB. Thankfully my displays have mounting brackets and I can wire them free hand. The center post will hold what ever I decide to use for the colon.




This is the tentative face plate. I have a design in mind that I need to refine. I'm thinking of going for a modern art-deco look. Some thing that has presence but isn't too flashy. The cut outs on the right are for AM/PM and alarm. I've gone though a few versions of these. I'll print out the newest ones tonight and show a few pictures tomorrow. With no data sheet it took a while to figure the right mounting spacing.

[Mysion]

As promised I printed out the parts so far and here they are!



The indicator window on the right doesn't quite line up but it's an easy fix. One thing I did notice was the printer was printing about 10 thou over. Not a huge issue but I have to be mindful. In the final design only the  sub frame will be 3D printed. The outer parts will be wood.


The design for the clock took me a while of just staring at the displays in the sub frame. Eventually I came up with one I liked. After fiddling around with CAD for a while hear it is!



Sorta of a modern art-deco look I think. I put a ruler at the bottom for scale. It's not massive but it's a good size. 8.5 inches wide. 4.5 inches tall and 3 in deep. I need to work on the hole in the back for the buttons next and try to find a really long 1/16 end mill.


PCB should come in this week so I have that to look forward to!

[Benta]

Nice!
I'd angle the hrs : mins separator as well, but that's just my taste.

[Mysion]

Thanks! That's not a bad point about the sperator. I modified it with an angle and here's the angled version



I like the consistency but I'm not so sure if I like how it's not evenly in the center any more. I'll have to think that one over.

[joshtyler]

OOOOO those displays are lovely... if you see a similar set on ebay please post the link.

There are these. The seller apparently has 10,000. I must say that I am tempted!

[SL4P]

I got away with three buttons for the UI.
Any button/function- timeout after 10 secs = return to clock mode.
BTN 1: Select mode - step across individual digits, then through various functions
BTN 2: increment (digit or displayed value)
BTN 3: decrement (digit or displayed value)

If the clock uses muxed LEDs etc, when not in 'digit set' mode - the inc/dec buttons adjust the brightness.

[tooki]

I got away with three buttons for the UI.
Any button/function- timeout after 10 secs = return to clock mode.
BTN 1: Select mode - step across individual digits, then through various functions
BTN 2: increment (digit or displayed value)
BTN 3: decrement (digit or displayed value)

If the clock uses muxed LEDs etc, when not in 'digit set' mode - the inc/dec buttons adjust the brightness.

Already massively better than all the Chinese clocks/kits that use two buttons.

With regards to alarm clocks, this is where (to the best of my knowledge), Sony got it right ages ago, and nobody else has copied: most Sony alarm clocks are designed such that verifying the wake time also turns that alarm on, completely eliminating the common problem of the user checking what time the alarm is set to, but forgetting to actually turn it on!! Sony alarm clocks with a single display do this by using a button for the alarm, which you hold to set the wake time, and that automatically turns on the alarm (a separate, much smaller button turns it off). In other models, they have a separate display for each wake time, with separate up/down buttons for each alarm, and a button that turns the alarm on and off -- and when the alarm is off, its display is also turned off, so again, checking the wake time necessarily requires turning it on!

It is the latter paradigm that I was planning to copy, probably using a little OLED display for the wake times.

[Mysion]

My biggest gripe with the damned cheap clocks is the ones with an Up botton but no down. To go from 36-34 you have to push it 58 god dam times. 
I might go full office space on it once I have a replacement.

Interesting idea from the Sony clocks. Might have to try it depending on how well mine works out. I figure it'll be one of those things were I'll constantly be tweaking it.

I'll be going for some thing similar to what you have SL4P just with 4.

I do have a small update. I picked out the wood at the local wood shop.

This is the wood I plan to use. It's sapele and ash. I think these two will complement each other nicely.



Here's a bonus pic of the very messy CNC. I need to do a bit of cleaning....   

[tooki]

My biggest gripe with the damned cheap clocks is the ones with an Up botton but no down. To go from 36-34 you have to push it 58 god dam times. 

And that's if you press "up" slowly. If you do it at a reasonable speed, the crappy debouncing code will make it miss half your presses…

Interesting idea from the Sony clocks. Might have to try it depending on how well mine works out. I figure it'll be one of those things were I'll constantly be tweaking it.

If you need demos of the Sony approaches, lemme know. My current alarm is the kind with separate clock displays, but my mom still has my old one with a single display.

[Gyro]

My biggest gripe with the damned cheap clocks is the ones with an Up botton but no down. To go from 36-34 you have to push it 58 god dam times. 
I might go full office space on it once I have a replacement.

Interesting idea from the Sony clocks. Might have to try it depending on how well mine works out. I figure it'll be one of those things were I'll constantly be tweaking it.

I'll be going for some thing similar to what you have SL4P just with 4.

I do have a small update. I picked out the wood at the local wood shop.

This is the wood I plan to use. It's sapele and ash. I think these two will complement each other nicely.



Here's a bonus pic of the very messy CNC. I need to do a bit of cleaning....   

Looks a nice contrast - are you planning on needing a few attempts?   

[Mysion]

Tooki thanks for the offer but I'll hold off right now. It may very well work good enough for my use.

I'm not planning on needing multiple attempts but it's a very real possibility, I've got way more wood then I need. I will probably do a short test run to see how well each wood machines and test what finish to use. I'll either finish it with clear or amber shellac.

My biggest worry is breaking an exotic 0.75 inch long 1/16 end mill. I've got two coming, hopefully I don't need more.

[SeanB]

Wood finish try sanding to 1200 grit then using some linseed oil or Danish oil to get a nice matt finish. I like all my oiled furniture, they have character to them from use, and as a bonus living in the borer belt they are almost immune to being eaten after they have had a number of coats applied to them. Walnut, Cherry and oak all come out nicely, and even cheap shutterply pine comes out looking reasonable with that.

I hate the modern particle board as that does get borer in them, because they stopped using formaldehyde based resins to bind the powder and chip together.

[Mysion]

If it all goes to plan I won't need to do to much sanding. I got an amazing finish on the CNC in the past. Sanding all depends on how well the edges line up when I laminate them. Hopefully I will be able to skip straight to fine sand paper. I'd consider oil but I already have shellac on hand.

Here's a few signs I've made in the past. No sanding done and finished with two light shellac coats. Indoors only due to the shellac.



The top is oak and the bottom one Mahogany. It's either high quality African mahogany or the genuine stuff. I haven't been able to find Mahogany since with a color that deep.

[Mysion]

The PCB's came today! Looking good! Project is starting to take shape.




The CAD design of the case is nearly done. I figured out how to have all the layers lock together for gluing. Took me a while. All that's left is the rear hole pattern for the buttons. I think I may have to wait until the rest of the parts come in. Some of the data sheets are lacking so I will have to take measurements.

Made a good start on the software too. With only 4K of program space it'll be fun! Honestly I don't anticipate any issues with only 4K.

[Mysion]

Well shit. The parts are in and they don't fit! Yay....  Turns out I used the pad out line for layout and not the much larger physical switch package.  I could bodge it but I have very tight physical limitation in the case. Plus I have a similar issue with the LED mounts as well. The lesson here is order your parts first and test fit them on a print out of the PCB. Lesson learned.



Well I'll make the best of this. I'll fit some of the circuitry and do a limited test of my driver. I should be able to remove most of the parts when the new board comes in. Maybe fix some of that wasted space too.

[Mysion]

Even though the PCB was flawed I decided to test what I could. I soldered on the buzzer, 12V power converter and part of the display drivers. The ATMega 48A was put on as well.



On testing the 12V supply worked fine. It floated at around 13.3V, a little high with load but for my uses it's good enough.  Buzzer had no issues thankfully. However my driver circuit had a minor error in it. Swapped a pad around in CAD and now the PNP MOSFETS will do what there supposed to.

I added a few minor things I forgot as well. Just a few debug test points. SDA, SCL and the like. 

Next up I'm getting ready to fire up the CNC. I've done a few test cuts. The Ash is very sensitive to heat and burns very easy. The Sapele looks like it will be a breeze to machine compared to the Ash.

[Mysion]

Alright time for a big update! It's CNC fun time.

Here we can see the CNC in action. It's cutting out one of the ash layers. It's a tedious process. I have to first flatten each side so he sticky tape can get a good grip. Then I have to get it to the right thickness before I start cutting the part out. Mistakes were made but it went well enough. I have to repeat this 6 times. The front will be the hardest so I'm saving it for last.


If there's interest I can get a video of the CNC in action.

There are 4 out of 6 layers done right now. I built in alignment posts into them. It all fits together very well, hopefully it will need minimal sanding. I really like how the colors are turning out so far. Feels really good to have it in my hand. I never heard of Sapele before this project and I'm loving this wood!




Last but not least the new PCB came in. You can see the new one on the bottom! You can see the different switches I decided to use with a few other minor changes. I went with a through hole buzzer too. The SMD one was hard to solder without melting it. 
The digikey shipment arrived at the same time. Perfect timing!

[Mr. Scram]

Those wooden parts look gorgeous, even though the sharp corners and protrusions across the grain scare me a bit. A hard knock or a screw that's too big might be all that is needed to crack the wood.

I really, really need to get my hands on a CNC machine.

[floobydust]

In firmware I suggest staggering the "flips" to lower peak current drain.
i.e "1"->"2" is 5 flips for one digit, and when digits roll over, there can be many to flip which can overload your power supply. Like 200msec per digit delay.

I did a outdoor flip-display billboard and used higher voltage with capacitive-discharge. Big flips caused the power to sag too much.

[Mysion]

Scram you make a good point.....  I may be guilty of form over function slightly. As long as I don't drop it though it should be good, with the exception of the screw hold not in a corner.
In retrospect I should have had the grain of the Ash 90 degrees relative to the Sapele. Ah well too late now... 

Yeah a CNC is a fun toy. I've done aluminium on mine okayish. It's a Sapeoko 3 XXL for reference.

Flooby dust you hit on an important point. My 12 supply is limited to 70 mA and each digit needs 20 mA for 20 mS to flip.
I'm not yet sure how I will flip them, if I use a pattern or not. I will have to sager them like you suggested. Thankfully there's not much of an inductive kick back if an when I looked on the scope. Better not be since I forgot protection diodes. 

[Mr. Scram]

Scram you make a good point.....  I may be guilty of form over function slightly. As long as I don't drop it though it should be good, with the exception of the screw hold not in a corner.
In retrospect I should have had the grain of the Ash 90 degrees relative to the Sapele. Ah well too late now... 

Yeah a CNC is a fun toy. I've done aluminium on mine okayish. It's a Sapeoko 3 XXL for reference.

Flooby dust you hit on an important point. My 12 supply is limited to 70 mA and each digit needs 20 mA for 20 mS to flip.
I'm not yet sure how I will flip them, if I use a pattern or not. I will have to sager them like you suggested. Thankfully there's not much of an inductive kick back if an when I looked on the scope. Better not be since I forgot protection diodes. 

Can you tell a little more about the CNC machine? Does it do brass and other softer metals too? Do you get any software with it, or do you need to supply that yourself?

I don't recall seeing a CNC machine with this type of spindle before. Usually it's one of those Dremel style ones, or just a full blown water cooled one.

[Mysion]

Sure just ask away!

I haven't done brass on mine yet. However if it can do aluminum it will be able to do brass. I haven't done any projects in aluminum yet but I do have a test piece I can show.



I used a 1/16 end mill and got good results. It's all within 10-20 thou if I recall correctly. It is a cnc router not mill and belt driven. Your not going to be making any thing really precise on it.
Still very capable and I have about a 1 meter by 1 meter routing area to play with. That said if you plan to work with metal all the time you need a proper mill.

As for the spindle it's a dewalt DW618 wood router. The CNC it's self came with the router and software to run gcode. All I had to make was the table. I have the one here https://shop.carbide3d.com/collections/frontpage/products/shapeoko-xl-kit?variant=42721938758
There is a smaller version too.

Now to generate the gcode I use the CAM feature of fusion 360. I do all my cad work and CNC code generation with it. As long as you make less than 100,000$ a year with it fusion 360 is free. (A problem I'd love to have)

Like I said if you have any more questions ask away!

[Mysion]

Ok I have good news and bad news. Good news is the PCB has been assembled and tested. It's working great. Bad news is the CNC is fucked. 
 
Here's a picture of the finished board. The 12V boost module is on the lower right. The RTC is hidden under the coin cell. The battery makes a great ground plane!



The good part is the CNC broke after I preformed multiple tool changes for the front plate. It would be a real shame to ruin it after so much work. You can see it with the epoxy in the inlay. The inlay is for the AM/PM and ALM indicator.
I was running the gcode for the colon insert in a block of derlin. It's the black object on the bottom right. Then the x axis locked up and the machine crashed.   



It's either the motor, cable, or driver. First failure I've had with the machine. Heck still more reliable than the 3D printer.

[Mysion]

Good news the CNC is fixed! Bad news is the control board needed to be replaced.

A stepper driver died on the dead board. I tried replacing it but it didn't work. I think the problem is it was too hot for to long. There is a ground plane in the board that acts as the heat sink. My hot air gun lacked the power to get the chip off and the chip was at high temperatures for a long time. Either that killed it or there was an internal trace damaged.

Next time I need a hot plate or oven. The new board is on the right in the picture below.



Once the CNC was fixed I finished off the case. The front indicator text was done using epoxy inlay. Even though it's not sanded yet it's looking real good! It looks exactly like what I expected with my cad model. I really love the contrast.





Now I just need to start sanding.

[Mysion]

After a lot of work the case has been sanded and then finished with shellac. I love the look of it! The sapele and ash really pops! I didn't think I had wood working in me!
I started with 80 grit sand paper and worked my way to 220 grit.

I also mounted the LED holder on the rear panel. It angles about 4 LED's 30 degrees from the vertical. I plan to activate them using a capacitive sensor on the top of the clock. Useful for the alarm too.



The letters on the rear pannel were made with a stencil. The stencil was cut out of 1/32 inch ply with a 1/32 inch end mill. The edges are blurry but it came out well enough. You can see a test piece with the stencil below.



With the case done I decided to start playing with the capacitive sensor. I just placed a short peice of copper tape on top of the display holder. It has to work through about 3/8 inch of wood.
It's pretty simple in operation. This micro controller drives it high then switches to a 10 meg-ohm resistor to drain it. Through testing I found out that it took slightly less than 150 uS for it to drain normally. If your within about 1/2 inch the sensor will trigger.



The next thing to do is to start wiring all the displays together and condensing it all down to a single 19 pin header. With some rough math I calculated about 210 soldering connections that I will have to make. Here's what I'm working with.



It's going to take a while... better keep working on it. 

[Mr. Scram]

After a lot of work the case has been sanded and then finished with shellac. I love the look of it! The sapele and ash really pops! I didn't think I had wood working in me!
I started with 80 grit sand paper and worked my way to 220 grit.

I also mounted the LED holder on the rear panel. It angles about 4 LED's 30 degrees from the vertical. I plan to activate them using a capacitive sensor on the top of the clock. Useful for the alarm too.



The letters on the rear pannel were made with a stencil. The stencil was cut out of 1/32 inch ply with a 1/32 inch end mill. The edges are blurry but it came out well enough. You can see a test piece with the stencil below.



With the case done I decided to start playing with the capacitive sensor. I just placed a short peice of copper tape on top of the display holder. It has to work through about 3/8 inch of wood.
It's pretty simple in operation. This micro controller drives it high then switches to a 10 meg-ohm resistor to drain it. Through testing I found out that it took slightly less than 150 uS for it to drain normally. If your within about 1/2 inch the sensor will trigger.



The next thing to do is to start wiring all the displays together and condensing it all down to a single 19 pin header. With some rough math I calculated about 210 soldering connections that I will have to make. Here's what I'm working with.



It's going to take a while... better keep working on it. 

A useful trick to prevent blurry edges, or underspray, is to use spray glue adhesive on the back. You let that dry until it's tacky and then spray paint your work piece. The glue seals the edges neatly, while also releasing after spraying. Generally, it says that you can reposition things on the can. You can get amazingly fine details that way, where you would normally get all sorts of messiness.

Can you tell a little bit more about the capacitive system and what chips or technology you used?

[Mysion]

That's a good tip with the glue! How much working time does it give you?
I'll have to give it a shot next time I need small text.


The capacitive part is surprisingly very simple. I'm not using any special chips just a 10M resistor and a copper sheet. Here's the circuit for the sensor.



(The 0J resistor is just a jumper)

The line going off screen connects to an IO on the AVR chip. The capacitive plate part is simply a header that goes to the larger copper area you see on the frame.
The AVR chip drives the line high. I use 1mS. Longer than needed just to make sure it's charged.

Then I switch the pin to high impedance mode and wait 150uS. At the end of 150uS I measure the value of the pin. If it reads high than some one has a hand near it.

They way it works if the plate acts as a capacitor similar to phone touch screen. It's discharges through a 10M resistor to ground. If you touch it or are near if you increase the capacitance of the plate.
Since humans act as earth referenced capacitors for some AC signals. The increased capacitance means that it's not discharged at the end of the 150uS time period and the chip will read a high.


I hope that clears it up. If any part of thats still fuzzy just ask.

[Mr. Scram]

That's a good tip with the glue! How much working time does it give you?
I'll have to give it a shot next time I need small text.


The capacitive part is surprisingly very simple. I'm not using any special chips just a 10M resistor and a copper sheet. Here's the circuit for the sensor.



(The 0J resistor is just a jumper)

The line going off screen connects to an IO on the AVR chip. The capacitive plate part is simply a header that goes to the larger copper area you see on the frame.
The AVR chip drives the line high. I use 1mS. Longer than needed just to make sure it's charged.

Then I switch the pin to high impedance mode and wait 150uS. At the end of 150uS I measure the value of the pin. If it reads high than some one has a hand near it.

They way it works if the plate acts as a capacitor similar to phone touch screen. It's discharges through a 10M resistor to ground. If you touch it or are near if you increase the capacitance of the plate.
Since humans act as earth referenced capacitors for some AC signals. The increased capacitance means that it's not discharged at the end of the 150uS time period and the chip will read a high.


I hope that clears it up. If any part of thats still fuzzy just ask.

That's an interesting and simple technique. I just know of various proprietary techniques, or at least proprietary names given to it. I might have to tinker around with this sometime. Even though a good clickety-clack switch can't be beat in terms of user feedback and zero power consumption, a solid state interface has its perks.

The working time of the glue doesn't seem to pose any problems. It's a rather weak glue meant to, maybe, stick permanently when applied right after spraying, or be temporary or "repositionable" after letting it dry for a bit. Various name and no name brands seem to carry something similar. I think it's intended to stick posters to foamcore boards and such. With a little bit of practice, it's quite easy to poke the glue and know whether it's good. It being too wet is quite obvious and it being totally dry takes a while. At worst, you can add another layer and get it started again. If done even roughly right, the backside of your stencil will be a bit like a post-it glue strip.

You can use the same stencil to paint multiple objects before the glue is fully dry and generally the front becoming too loaded with paint is more of a problem than the back drying out. Of course, one drawback is that reapplying paint on the back when the front is covered in paint can be a bit problematic, but if you let the stencil dry out completely, it should be reusable again, other than possible going wrinkly on you.

It's a little trick I picked up years ago from people who do complex street art style multi layered stencil paintings. It has served me well over the years in all sorts of projects.

https://www.graphicsdirect.co.uk/3m-spraymount-400ml.html

[Mysion]

Well after much waiting the new display adapter boards came in. This time each enable or disable pin has a diode preventing back feeding. I also redid the wiring and went with ribbon cables this time. The hand wiring was tedious to say the least and unreliable.



With a proper crimping tool ribbon cables are a breeze. The two cuts you see in the cable allow it to be routed along the support posts on the frame. Then the lower cable simply attaches on and breaks out the connections to the main board. Pretty slick.

With the board wired up properly time to work on the display code! Finally I get to see this thing tick. For the code I used an event loop based system. The clock iterates though a loop every 10-20 mS. This might not be the most elegant way but it worked well for the features I wanted. For most of the buttons holding them down triggers and alternate function, the event loops lets me easily keep track of how long every button has been held down. However that was a major gotcha that I ran into that almost derailed the project. I almost ran out of space in the 4096 byte chip. I ended up doing some serious code optimizations. I came in at 4084 bytes used out of 4096. I did get all the features in that I wanted so I just barely squeaked by!



Alright I'm getting close to the end here! Time for final assembly!  One of my design goals was to have the project breakdown easily. Judging by the picture below I'd say I accomplished this goal.



Now I just have to made a few final touched and make sure it all fits. I'm getting real close to calling it done, I should have the final post up real soon.